⚛️ quantum physics

Towards self-correcting quantum codes for neutral atom arrays

This paper introduces "ZSZ codes," a non-abelian generalization of bivariate bicycle codes tailored for neutral atom arrays, which demonstrate competitive performance with standard decoders and superior sustainable thresholds under local self-correcting decoders, positioning them as promising candidates for scalable fault-tolerant quantum memories.

Original authors: Jinkang Guo, Yifan Hong, Adam Kaufman, Andrew Lucas

Published 2026-01-15

📖 5 min read🧠 Deep dive

CC BY 4.0

Original authors: Jinkang Guo, Yifan Hong, Adam Kaufman, Andrew Lucas

Original paper licensed under CC BY 4.0 (http://creativecommons.org/licenses/by/4.0/). ✨ This is an AI-generated explanation of the paper below. It is not written or endorsed by the authors. For technical accuracy, refer to the original paper. Read full disclaimer

Imagine you are trying to store a precious secret in a room full of mischievous gremlins. These gremlins represent noise in a quantum computer—they constantly try to flip switches, change values, and corrupt your information. To fight them, scientists use Quantum Error Correction (QEC). Think of this as a team of security guards (the "code") constantly checking the room to see if anything has changed. If they find a mistake, they fix it before the secret is lost.

For a long time, the best security teams were like Surface Codes. They are reliable, but they are very "expensive" in terms of space. You need a huge number of physical guards (qubits) to protect just a few secrets (logical qubits). It's like hiring a thousand security guards to watch a single bank vault.

Recently, scientists discovered a more efficient team called Bivariate Bicycle (BB) codes. They are like a leaner, faster security squad that uses fewer guards but still does a great job. However, they have a weakness: they can't easily "self-correct." If the guards get confused or the noise gets too heavy, they need a central commander (a complex computer) to tell them what to do. This takes time and energy.

This paper introduces a new, improved security team called ZSZ codes. Here is how they work, explained simply:

1. The "Twisted" Layout

The old BB codes are built on a flat, grid-like floor plan where everyone follows simple, straight rules (like a chessboard). The new ZSZ codes are built on a "twisted" floor plan.

Imagine a hotel where the hallways don't just go straight; they loop around in a special, non-Euclidean way. If you walk down a hallway and turn a corner, you might end up in a different part of the building than you expected. This "twist" is a mathematical trick called a semidirect product. It sounds complicated, but the result is that the security guards are connected in a much more complex, web-like network.

2. The "Self-Correcting" Superpower

The biggest breakthrough with ZSZ codes is self-correction.

Old Way (BB Codes): If a guard sees a mistake, they have to shout it out to a central computer. The computer calculates the fix and tells the guard what to do. This takes time. If the noise is too loud, the system crashes.
New Way (ZSZ Codes): Because of the twisted layout, the guards are so interconnected that if one guard sees a mistake, the local rules of the building naturally push the error away. It's like a crowd of people in a hallway: if someone tries to push through, the natural flow of the crowd pushes them back without needing a manager to intervene.

The paper calls this passive error correction. The system fixes itself automatically, like a thermostat that turns on the heat when the room gets cold, without you having to touch the dial.

3. The Results: A Stronger Shield

The authors ran computer simulations to see how well these new codes hold up against the "gremlins" (noise).

The Threshold: They found that ZSZ codes can handle a noise level of about 0.095% before they start failing.
The Comparison: This is significantly better than the 4D Toric code (another famous self-correcting code), which fails at around 0.06%.
The Takeaway: ZSZ codes are more robust. They can survive in a "noisier" environment than previous self-correcting codes, making them a much better candidate for building a long-lasting quantum memory.

4. How to Build It: The "Moving Furniture" Analogy

You might wonder, "How do you build a twisted, non-flat floor plan in a real quantum computer?"

The paper suggests using neutral atom arrays. Imagine a grid of tiny traps holding individual atoms (the qubits). Usually, these atoms are stuck in place. But in this setup, scientists use optical tweezers (lasers that act like invisible fingers) to pick up atoms and move them around.

To perform the "security checks" (syndrome extraction), the researchers propose a dance routine:

They pick up rows and columns of atoms.
They slide them around in a "riffle shuffle" (like shuffling a deck of cards).
They bring them together to check for errors.
They slide them back to their original spots.

Because the atoms can move, they can create the complex, twisted connections required by the ZSZ code, even though the physical hardware is just a flat grid of lasers.

Summary

The paper proposes a new type of quantum error-correcting code called ZSZ codes.

What it is: A mathematically "twisted" version of existing efficient codes.
Why it matters: It allows the quantum computer to self-correct errors automatically without needing constant external intervention.
The Proof: Simulations show it has a higher "survival threshold" (can handle more noise) than previous self-correcting codes.
The Hardware: It can be built using neutral atoms that are physically moved around by lasers to create the necessary connections.

In short, the authors have found a way to make quantum memory more robust and self-sufficient, potentially paving the way for quantum computers that can run for longer without crashing.

Technical Summary: Towards Self-Correcting Quantum Codes for Neutral Atom Arrays

1. Problem Statement

The development of fault-tolerant quantum computation requires low-overhead quantum error-correcting codes (QECCs). While surface codes are widely used, they suffer from high physical qubit overhead. Bivariate bicycle (BB) codes have emerged as a promising alternative, offering competitive performance with significantly fewer physical qubits by utilizing non-local connectivity. However, BB codes are constrained by finite-dimensional Euclidean embeddings, which limit their ultimate scaling properties.

A critical challenge in quantum memory is the implementation of "passive" or "self-correcting" error correction. Traditional active decoding requires mid-circuit measurements and feedback, introducing latency and complexity. Passive decoding, inspired by self-correcting memories, utilizes local, constant-depth circuits (often measurement-free) to continuously suppress errors. While 4D toric codes exhibit self-correction, they require high-dimensional connectivity that is difficult to realize physically. The authors investigate whether there exist quantum Low-Density Parity-Check (LDPC) codes that combine the low overhead of BB codes with the potential for self-correction, specifically tailored for neutral atom platforms capable of long-range connectivity.

2. Methodology

2.1. Code Construction: ZSZ Codes

The authors introduce ZSZ codes, a family of quantum LDPC codes defined as a non-abelian generalization of BB codes.

Algebraic Foundation: The codes are constructed using the Two-Block Group Algebra (2BGA) framework. They utilize the semidirect product group $G = \mathbb{Z}_\ell \rtimes_q \mathbb{Z}_m$ , defined by the presentation $\langle x, y \mid x^\ell = y^m = yxy^{-1}x^{-q} = 1 \rangle$ , where $qm \equiv 1 \pmod \ell$ .
Twist Mechanism: Unlike BB codes (which correspond to the abelian case $q=1$ ), ZSZ codes employ a "twist" (semidirect product) where the action of $y$ on $x$ is non-trivial ( $yxy^{-1} = x^q$ ). This modifies the microscopic connection rules between qubits and parity checks.
Structure: The codes are CSS codes where each parity check involves six qubits, and each qubit participates in six checks (three X-type, three Z-type). The parity check matrices $H_X$ and $H_Z$ are constructed from the left-regular and right-regular representations of group algebra elements $a, b \in \mathbb{F}_2[G]$ .
Geometry: The codes admit a 2D rectangular embedding where qubits reside on the links of an $\ell \times m$ lattice with periodic boundaries. The non-abelian twist alters the connectivity, effectively creating a Cayley graph structure that differs from the standard rectangular lattice of BB codes.

2.2. Numerical Simulations

The authors performed extensive numerical simulations under a circuit-level depolarizing noise model to evaluate code performance using two distinct decoding strategies:

Global Decoding (Active): Simulating $d$ rounds of syndrome extraction followed by a global Belief-Propagation and Ordered-Statistics Decoder (BP+OSD).
Passive Decoding (Self-Correcting): Simulating $\ge 100$ rounds of syndrome extraction using a local, single-shot "greedy" decoder. This decoder acts as a zero-temperature Gibbs sampler, applying local corrections to minimize syndrome weight without mid-circuit measurements.

2.3. Hardware Implementation Proposal

The paper proposes a specific implementation strategy for ZSZ codes on neutral atom arrays trapped in movable tweezer arrays.

Syndrome Extraction: A single-ancilla syndrome extraction protocol is designed where a syndrome qubit is coupled to data qubits via optical tweezers.
Routing: The authors detail a routing protocol using Acousto-Optic Deflectors (AODs) to move atoms. The complexity is $O(\log \ell)$ for the horizontal sector (left Cayley graph routing) and $O(m)$ for the vertical sector (right Cayley graph routing). For codes with logarithmic diameter ( $m = O(\log \ell)$ ), the total routing complexity is $O(\log \ell)$ .

3. Key Contributions and Results

3.1. Code Properties

Expansion and Confinement: Theoretical analysis suggests that ZSZ codes exhibit small-set expansion in their Cayley graphs, leading to linear confinement for small error sets. This is a necessary condition for self-correction, contrasting with BB codes which do not exhibit this property in the same regime.
Parameters: The authors identified several candidate codes (e.g., ZSZ144, ZSZ288, ZSZ756) with varying numbers of logical qubits ( $k$ ) and estimated distances ( $d$ ).

3.2. Performance under Global Decoding

Under $d$ rounds of syndrome extraction with BP+OSD decoding, ZSZ codes demonstrate a threshold of approximately 0.5% under circuit-level depolarizing noise.
This performance is competitive with the surface code (estimated threshold ~0.8% under the same model) and BB codes, while encoding significantly more logical qubits for a given physical size.

3.3. Performance under Passive Decoding

Sustainable Threshold: Using the local greedy decoder, the authors observe a sustainable threshold of approximately 0.095% for ZSZ codes.
Comparison: This threshold is notably higher than the estimated 0.06% sustainable threshold for the 4D toric code under the same noise model and decoder.
BB Code Contrast: Crucially, the authors report no evidence of a sustainable threshold for BB codes under the same passive decoding scheme, suggesting that the non-abelian twist in ZSZ codes fundamentally alters the error correction dynamics in a way that favors passive correction.
Resource Efficiency: A ZSZ756 code outperforms four copies of a $L=4$ 4D toric code while using only ~12% of the data qubits for the same number of logical qubits.

3.4. Implementation Feasibility

The paper demonstrates that a complete round of syndrome extraction for ZSZ codes can be achieved using simple global motions of atomic arrays (grid transfers and cyclic shifts). While the routing complexity ( $O(\log \ell)$ ) is higher than that of BB codes ( $O(1)$ ), it remains logarithmic and comparable to hypergraph product codes, making it feasible for neutral atom architectures.

4. Significance and Claims

The paper claims that ZSZ codes represent a promising candidate for scalable self-correcting quantum memories. The primary significance lies in the demonstration that a simple non-abelian generalization of BB codes can yield:

Superior Passive Performance: A sustainable threshold for passive decoding that exceeds that of the 4D toric code, a benchmark for self-correction.
Hardware Compatibility: A viable path to implementation in neutral atom platforms, which are uniquely suited to handle the non-local connectivity required by these codes.
New Paradigm: The results suggest that non-abelian group structures can be leveraged to design quantum codes with qualitative behaviors (like self-correction) that are absent in their abelian counterparts (BB codes), even at moderate code lengths.

The authors remain modest regarding the immediate practicality, noting that the observed thresholds are based on specific noise models and decoders. They acknowledge that further work is needed to optimize syndrome extraction scheduling, mitigate hook errors, and develop fault-tolerant state preparation and logical gate protocols for these codes. However, the numerical evidence of a sustainable threshold higher than the 4D toric code establishes ZSZ codes as a significant step toward measurement-free, scalable quantum memory.